MASTERS PROJECT REPORT Cache Consistency Techniques for Peer-to-Peer File Sharing Networks

Peer-to-peer (P2P) file sharing systems provide infrastructure for communities to share storage space (e.g., Napster, Gnutella [1], Kazza [2]). Unlike traditional distributed systems, P2P networks aim to aggregate large numbers of computers that join and leave the network frequently and might not have permanent network (IP) addresses. In a pure distributed P2P system such as Gnutella, peers communicate directly with each other and share resources without dedicated servers. In essense, these P2P systems build, at the application level, a virtual overlay network with it own routing mechanisms. Popular usage of P2P applications has focused simple sharing of music or video files. However, P2P file sharing systems are not limited to this task. Because P2P networks are application-level networks, new functionalities can be easily added and deployed. For example, the Gnutella protocol was designed to share any type of file, and new features have been continuously added or proposed to the existing model. These observations have motivated us to consider cache consistency issues in a P2P system, like Gnutella. Using the web model as a comparison, we assume that each file in a Gnutella network has an origin server. And correspondingly, the file at its origin server is called its master copy. Any user in the network can download any file she finds through querying, and thus, files are replicated by user initiatives. We consider all the non-master-copies of a file as replica of the master copy, and assume that only the master copy of a file can be modified. Given this terminology, maintaining cache consistency in the Gnutella